The first EU agency approved, the EU Network Security Agency will adopt the Anthropic Mythos model.

Anthropic PBC is about to allow the European Union Agency for Cybersecurity (ENISA) access to its powerful artificial intelligence tool called "Mythos," making it the first EU agency to have access to this powerful AI tool. The tool has the capability to discover and exploit vulnerabilities in computer systems. According to sources, the generative AI company will allow ENISA to join the "Project Glasswing." This project aims to allow key institutions to test the functionality of Mythos before it is more widely released. Sources revealed that Anthropic informed the European Commission of this decision over the weekend. Prior to this, European Commission officials had traveled to San Francisco last Thursday to request access to the model from Anthropic, but previous attempts had been unsuccessful. A spokesperson for the European Commission stated that discussions with Anthropic are progressing but declined to comment further. Since its first preview in April of this year, access to Mythos has been restricted due to concerns that it could fall into the hands of cybercriminals. Anthropic has issued warnings that Mythos can automatically discover and exploit high-risk vulnerabilities in mainstream operating systems and browsers, autonomously developing sophisticated cyber attack capabilities. If in the wrong hands, it could have a serious impact on global digital infrastructure, public safety, and national security. US authorities, as well as some US financial institutions including Goldman Sachs, have already been granted early access. The UK's Artificial Intelligence Security Institute was also granted testing access shortly after the model was launched. Global financial institutions are urgently responding to Mythos. Concerns among global financial institutions about Mythos are focused on the unequal distribution of defense resources. In order to manage risks, Anthropic previously initiated a closed project called "Project Glasswing," providing restricted access to about 50 selected US technology and financial giants such as Google, Microsoft, and JPMorgan Chase, aiming to use the model for defensive cybersecurity research and strengthen critical infrastructure in advance. This measure previously prevented European banks and regulatory agencies from gaining access to Mythos, causing widespread anxiety. Frank Elderson, Vice-Chair of the Supervisory Board of the European Central Bank, admitted that European banks not being able to access the model is "regrettable," but he emphasized that "not having access to the model is not an excuse for inaction" and hoped that US banks attending the meeting could share their testing experience with their Eurozone counterparts. On May 26, the European Central Bank convened a meeting with around 111 large banks in its jurisdiction to discuss the cybersecurity risks exposed by advanced AI models such as Mythos. Elderson pointed out that as AI technology advances, traditional security vulnerabilities must be addressed more quickly. If a major software vendor releases a security patch, hackers used to take weeks to reverse engineer the vulnerability, but with AI models like Mythos, malicious actors could potentially complete this process in just 30 minutes. Therefore, he demanded that banks establish new processes to ensure software patches are deployed and updated at a speed far exceeding current market practices. Currently, numerous institutions worldwide are applying to Anthropic for access to Mythos or related technical briefings. Anthropic has agreed to provide senior briefings to the Financial Stability Board comprising G20 finance ministers and central bank officials, as well as the European Commission.